src/compiler/shader_info.h

   1 /*
   2  * Copyright © 2016 Intel Corporation
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  21  * IN THE SOFTWARE.
  22  *
  23  */
  24
  25 #ifndef SHADER_INFO_H
  26 #define SHADER_INFO_H
  27
  28 #include "shader_enums.h"
  29 #include <stdint.h>
  30
  31 #ifdef __cplusplus
  32 extern "C" {
  33 #endif
  34
  35 struct spirv_supported_capabilities {
  36    bool address;
  37    bool atomic_storage;
  38    bool demote_to_helper_invocation;
  39    bool derivative_group;
  40    bool descriptor_array_dynamic_indexing;
  41    bool descriptor_array_non_uniform_indexing;
  42    bool descriptor_indexing;
  43    bool device_group;
  44    bool draw_parameters;
  45    bool float64;
  46    bool fragment_shader_sample_interlock;
  47    bool fragment_shader_pixel_interlock;
  48    bool geometry_streams;
  49    bool image_ms_array;
  50    bool image_read_without_format;
  51    bool image_write_without_format;
  52    bool int8;
  53    bool int16;
  54    bool int64;
  55    bool int64_atomics;
  56    bool kernel;
  57    bool min_lod;
  58    bool multiview;
  59    bool physical_storage_buffer_address;
  60    bool post_depth_coverage;
  61    bool runtime_descriptor_array;
  62    bool float_controls;
  63    bool shader_clock;
  64    bool shader_viewport_index_layer;
  65    bool stencil_export;
  66    bool storage_8bit;
  67    bool storage_16bit;
  68    bool storage_image_ms;
  69    bool subgroup_arithmetic;
  70    bool subgroup_ballot;
  71    bool subgroup_basic;
  72    bool subgroup_quad;
  73    bool subgroup_shuffle;
  74    bool subgroup_vote;
  75    bool tessellation;
  76    bool transform_feedback;
  77    bool variable_pointers;
  78    bool vk_memory_model;
  79    bool vk_memory_model_device_scope;
  80    bool float16;
  81    bool amd_gcn_shader;
  82    bool amd_shader_ballot;
  83    bool amd_trinary_minmax;
  84    bool amd_image_read_write_lod;
  85 };
  86
  87 typedef struct shader_info {
  88    const char *name;
  89
  90    /* Descriptive name provided by the client; may be NULL */
  91    const char *label;
  92
  93    /** The shader stage, such as MESA_SHADER_VERTEX. */
  94    gl_shader_stage stage:8;
  95
  96    /** The shader stage in a non SSO linked program that follows this stage,
  97      * such as MESA_SHADER_FRAGMENT.
  98      */
  99    gl_shader_stage next_stage:8;
 100
 101    /* Number of textures used by this shader */
 102    uint8_t num_textures;
 103    /* Number of uniform buffers used by this shader */
 104    uint8_t num_ubos;
 105    /* Number of atomic buffers used by this shader */
 106    uint8_t num_abos;
 107    /* Number of shader storage buffers (max .driver_location + 1) used by this
 108     * shader.  In the case of nir_lower_atomics_to_ssbo being used, this will
 109     * be the number of actual SSBOs in gl_program->info, and the lowered SSBOs
 110     * and atomic counters in nir_shader->info.
 111     */
 112    uint8_t num_ssbos;
 113    /* Number of images used by this shader */
 114    uint8_t num_images;
 115    /* Index of the last MSAA image. */
 116    int8_t last_msaa_image;
 117
 118    /* Which inputs are actually read */
 119    uint64_t inputs_read;
 120    /* Which outputs are actually written */
 121    uint64_t outputs_written;
 122    /* Which outputs are actually read */
 123    uint64_t outputs_read;
 124    /* Which system values are actually read */
 125    uint64_t system_values_read;
 126
 127    /* Which patch inputs are actually read */
 128    uint32_t patch_inputs_read;
 129    /* Which patch outputs are actually written */
 130    uint32_t patch_outputs_written;
 131    /* Which patch outputs are read */
 132    uint32_t patch_outputs_read;
 133
 134    /** Bitfield of which textures are used */
 135    uint32_t textures_used;
 136
 137    /** Bitfield of which textures are used by texelFetch() */
 138    uint32_t textures_used_by_txf;
 139
 140    /* SPV_KHR_float_controls: execution mode for floating point ops */
 141    uint16_t float_controls_execution_mode;
 142
 143    /* The size of the gl_ClipDistance[] array, if declared. */
 144    uint8_t clip_distance_array_size:4;
 145
 146    /* The size of the gl_CullDistance[] array, if declared. */
 147    uint8_t cull_distance_array_size:4;
 148
 149    /* Whether or not this shader ever uses textureGather() */
 150    bool uses_texture_gather:1;
 151
 152    /**
 153     * True if this shader uses the fddx/fddy opcodes.
 154     *
 155     * Note that this does not include the "fine" and "coarse" variants.
 156     */
 157    bool uses_fddx_fddy:1;
 158
 159    /**
 160     * True if this shader uses 64-bit ALU operations
 161     */
 162    bool uses_64bit:1;
 163
 164    /* Whether the first UBO is the default uniform buffer, i.e. uniforms. */
 165    bool first_ubo_is_default_ubo:1;
 166
 167    /* Whether or not separate shader objects were used */
 168    bool separate_shader:1;
 169
 170    /** Was this shader linked with any transform feedback varyings? */
 171    bool has_transform_feedback_varyings:1;
 172
 173    /* Whether flrp has been lowered. */
 174    bool flrp_lowered:1;
 175
 176    union {
 177       struct {
 178          /* Which inputs are doubles */
 179          uint64_t double_inputs;
 180
 181          /* For AMD-specific driver-internal shaders. It replaces vertex
 182           * buffer loads with code generating VS inputs from scalar registers.
 183           *
 184           * Valid values: SI_VS_BLIT_SGPRS_POS_*
 185           */
 186          uint8_t blit_sgprs_amd:4;
 187
 188          /* True if the shader writes position in window space coordinates pre-transform */
 189          bool window_space_position:1;
 190       } vs;
 191
 192       struct {
 193          /** The output primitive type (GL enum value) */
 194          uint16_t output_primitive;
 195
 196          /** The input primitive type (GL enum value) */
 197          uint16_t input_primitive;
 198
 199          /** The maximum number of vertices the geometry shader might write. */
 200          uint16_t vertices_out;
 201
 202          /** 1 .. MAX_GEOMETRY_SHADER_INVOCATIONS */
 203          uint8_t invocations;
 204
 205          /** The number of vertices recieves per input primitive (max. 6) */
 206          uint8_t vertices_in:3;
 207
 208          /** Whether or not this shader uses EndPrimitive */
 209          bool uses_end_primitive:1;
 210
 211          /** Whether or not this shader uses non-zero streams */
 212          bool uses_streams:1;
 213       } gs;
 214
 215       struct {
 216          bool uses_discard:1;
 217
 218          /**
 219           * True if this fragment shader requires helper invocations.  This
 220           * can be caused by the use of ALU derivative ops, texture
 221           * instructions which do implicit derivatives, and the use of quad
 222           * subgroup operations.
 223           */
 224          bool needs_helper_invocations:1;
 225
 226          /**
 227           * Whether any inputs are declared with the "sample" qualifier.
 228           */
 229          bool uses_sample_qualifier:1;
 230
 231          /**
 232           * Whether early fragment tests are enabled as defined by
 233           * ARB_shader_image_load_store.
 234           */
 235          bool early_fragment_tests:1;
 236
 237          /**
 238           * Defined by INTEL_conservative_rasterization.
 239           */
 240          bool inner_coverage:1;
 241
 242          bool post_depth_coverage:1;
 243
 244          /**
 245           * \name ARB_fragment_coord_conventions
 246           * @{
 247           */
 248          bool pixel_center_integer:1;
 249          bool origin_upper_left:1;
 250          /*@}*/
 251
 252          bool pixel_interlock_ordered:1;
 253          bool pixel_interlock_unordered:1;
 254          bool sample_interlock_ordered:1;
 255          bool sample_interlock_unordered:1;
 256
 257          /**
 258           * Flags whether NIR's base types on the FS color outputs should be
 259           * ignored.
 260           *
 261           * GLSL requires that fragment shader output base types match the
 262           * render target's base types for the behavior to be defined.  From
 263           * the GL 4.6 spec:
 264           *
 265           *     "If the values written by the fragment shader do not match the
 266           *      format(s) of the corresponding color buffer(s), the result is
 267           *      undefined."
 268           *
 269           * However, for NIR shaders translated from TGSI, we don't have the
 270           * output types any more, so the driver will need to do whatever
 271           * fixups are necessary to handle effectively untyped data being
 272           * output from the FS.
 273           */
 274          bool untyped_color_outputs:1;
 275
 276          /** gl_FragDepth layout for ARB_conservative_depth. */
 277          enum gl_frag_depth_layout depth_layout:3;
 278       } fs;
 279
 280       struct {
 281          uint16_t local_size[3];
 282
 283          bool local_size_variable:1;
 284          uint8_t user_data_components_amd:3;
 285
 286          /*
 287           * Arrangement of invocations used to calculate derivatives in a compute
 288           * shader.  From NV_compute_shader_derivatives.
 289           */
 290          enum gl_derivative_group derivative_group:2;
 291
 292          /**
 293           * Size of shared variables accessed by the compute shader.
 294           */
 295          unsigned shared_size;
 296
 297          /**
 298           * pointer size is:
 299           *   AddressingModelLogical:    0    (default)
 300           *   AddressingModelPhysical32: 32
 301           *   AddressingModelPhysical64: 64
 302           */
 303          unsigned ptr_size;
 304       } cs;
 305
 306       /* Applies to both TCS and TES. */
 307       struct {
 308          uint16_t primitive_mode; /* GL_TRIANGLES, GL_QUADS or GL_ISOLINES */
 309
 310          /** The number of vertices in the TCS output patch. */
 311          uint8_t tcs_vertices_out;
 312          enum gl_tess_spacing spacing:2;
 313
 314          /** Is the vertex order counterclockwise? */
 315          bool ccw:1;
 316          bool point_mode:1;
 317       } tess;
 318    };
 319 } shader_info;
 320
 321 #ifdef __cplusplus
 322 }
 323 #endif
 324
 325 #endif /* SHADER_INFO_H */